I wrote:
> I do agree that improving the names generated for expression indexes
> would be useful independently of this. I propose setting this patch
> aside for the time being, and I will go look into that, and then
> if that gets accepted we can come back here and discuss how much
> of a problem remains.
I started a separate thread for that, mostly so the cfbot doesn't
see it as the same patch:
https://www.postgresql.org/message-id/flat/876799.1757987810%40sss.pgh.pa.us
regards, tom lane